Running longer stress-tests against the qpid-cpp broker causes either the client (perftest) or the broker to crash with a glibc error:
invalid fastbin entry (free)
If a longer test is used, the probability of a crash is close to certain. The "invalid fastbin entry (!prev)" message is also seen, as are memory corruption messages.
Why this should be showing up only on RHEL-6 (and rawhide) is not clear.
The following test shows up the problem with reasonable consistency:
perftest --iterations 1 --summary --unique-data yes --log-enable info+ --mode topic --qt 1 --durable yes --npubs 2 --nsubs 3 --count 200000 --size 1024 --tx 0 --async-commit no
(perftest is not part of the core RHEL-6 packages, it is in the -devel packages in the MRG channel).
I recommend this be set to blocker status. (I cannot set that myself).
Created attachment 414914 [details]
I suspect there is a glibc problem. The attached test program uses no qpid code at all and exhibits the same symptom. As far as I can tell it has no errors in it.
Created attachment 414915 [details]
A similar but smaller test case
This requires the -std=c++0x option on compilation; it replaces the trivial wrappers around pthreads used in the previous test code with the std equivalents. It also exhibits the same symptom after a very short period (usually < 1 minute).
Fixed in glibc-2.12-1.1.el6.1
*** Bug 587694 has been marked as a duplicate of this bug. ***
Red Hat Enterprise Linux 6.0 is now available and should resolve
the problem described in this bug report. This report is therefore being closed
with a resolution of CURRENTRELEASE. You may reopen this bug report if the
solution does not work for you.